Use and Meaning

Richard G. Heck, Jr.

1 Opening
531

Many philosophers have found themselves tempted by the view that meaning is, in some way or other, determined by usewhich claim I shall call the Use-Meaning Thesis. Stated in such general terms, of course, the Thesis is merely programmatic. Until it is said what use is supposed to bethat is, in what terms use is to be characterizedand how use is supposed to determine meaning, it can function at most as a framework principle. The Thesis will thus admit of a wide range of specications, depending upon how the notion of meaning is explained and how use is thought to determine meaning, as well as upon how the notion of use itself is understood. My purpose in the present paper is to begin to clarify the Thesis by focusing on this last issue and, more specically, upon the questions what different characterizations of use might be available and how a choice among them might be made. I intend to approach these questions by considering John McDowells claim1 that anyone who would attempt to develop the Use-Meaning Thesis will face a dilemma when forced to answer the question whether use encompasses the contents expressed by utterances. If it does, then the right way to describe the use of a sentence will be in terms of what it can be used to say: the sentence Snow is white, for example, is used to express the thought that snow is white. But then, although the Thesis will surely be truethe meaning of a sentence can be taken to be the thought it is used to expressit will not be particularly likely to have far-reaching philosophical consequences. On the other hand, if the notion of use is not, in this sense, meaning-laden, it is hard to see how use should be explained, if the use | of a sentence is not to be characterized in broadly Quinean terms, that 532 is, in terms of the noises people make and the circumstances under which they make them. But then, the Use-Meaning Thesis will commit us to a behavioristic reduction of meaning, which few nowadays would nd appealing. The challenge McDowell poses, and which I want here to take up, is thus to explain how the
1 I shall draw chiey on McDowell (1998b). Similar views are expressed in a number of McDowells writings, to some of which I shall refer below.

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

Thesis can be both substantial and plausible, that is, to identify a notion of use that is non-behavioristic but not meaning-laden.

Use as Meaning: Metaphysical, Not Epistemological

No-one has done more to contribute to the development of the Use-Meaning Thesis than Michael Dummett, and it is with Dummetts conception of it, and its significance, that McDowells reections begin. According to McDowell (1998b, pp. 239ff), Dummetts refusal to countenance a meaning-laden notion of use is, at bottom, driven by a awed epistemology of understanding. On this picture, when someone makes an utterance, what are immediately available to other speakers are simply facts about what sounds were produced under what circumstances: The epistemological problem facing speakers is to reconstruct from this meager data the meanings carried by the sounds. One might call this picture of understanding the sense-datum conception of understanding, for it is, as McDowell emphasizes, analogous to the picture of perception found in sense-datum epistemologies: In that case, what are immediately available to me are but facts about my own experiential states, and my problem is to determine what my having these states might tell me about the external world. It is the corresponding picture, in the case of understanding, that is supposed to motivate Dummetts own conception of use, which must now be the Quinean one mentioned above: The use someone makes of a sentence is to be characterized behavioristically, in terms of the sounds she produces and the conditions under which she does soin terms that do not involve the notion of meaning, nor any psychological notions at all. For the problem is to show how one can get from the data immediately available in experience, itself conceived along broadly empiricist lines, to knowledge of what is said.2 It is worth emphasizing, before we move on, that the analogy between understanding and perception cuts both ways. McDowell wants to argue, in the perceptual case, that once we recognize that we literally see, say, that there is a lamp on the deskthat experience provides us with such conceptual contentswe should no longer concern ourselves with questions about what determines perceptual content.3 McDowell seems to think that one can only take seriously the question what
That there is something wrong with McDowells interpretation of Dummett is suggested by the fact that Dummett frequently expresses dissatisfaction with Quines conception of use. See, for example, his (1993d, p. 105), where he writes that the notion of knowledge cannot. . . be extruded from the philosophy of language. For references to others with similar interpretations of Dummett, and criticism of that reading, see Shieh (1998). 3 At least this seems to be part of the force of the argument of McDowell (1996): see lecture 3 and the appendix to it. I have in mind McDowells criticisms of Christopher Peacockes use of the notion of non-conceptual content in Peacocke (1992). For a little more on this, see note 38.
2

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

it is for particular experiences to have the contents they do if one thinks that experience as such, as it is in | itself, is content-free: The problem then becomes to 533 explain how we, as perceivers, can assign content to experiences that, in and of themselves, are without it, are mere subjective happenings. But the question what determines perceptual content is independent of the question whether experience has content, as it is in itself: We could, so far as I can see, even concede that an experience could not be the very experience it was without having the very content it did, and yet raise the question what determined its content, or what determined that someone had an experience with that content rather than some other. The present discussion of the case of understanding may therefore teach us lessons that can also be applied to the case of perception.4 McDowell argues in detail that the sense-datum conception of understanding is untenable, and I have no wish to quarrel with him. It is obviously incorrect, phenomenologically speaking, that we simply hear other speakers as making noises: We literally hear them as having said certain things, for example, that snow is white.5 And if the only reason to insist upon a meaning-free description of use were epistemological prejudice, that conception of use would have to be rejected. But there is an answer to this argument of McDowells: That a rejection of meaning-laden notions of use need not need be founded upon any epistemological doctrine, prejudicial or otherwise, but should rest upon a conception of the purpose of the Use-Meaning Thesis. McDowells discussion makes it seem as if the Thesis is a doctrine about how we, as speakers, are able to determine what others mean. But, properly understood, it proposes a framework for answering the question what it is for expressions to mean what they do, what determines what they mean, in a metaphysical sense. That is to say, the Use-Meaning Thesis is not an epistemological doctrine, but a metaphysical one.6 So the appropriateness of a particular conception of use for specifying the Thesis needs to be evaluated in terms of the metaphysical purposes for which it is wanted. We are not free to select any conception we please, however phenomenologically or epistemologically appealing it might be. To characterize use in terms that incorporate the notion of what a sentence means is to trivialize the metaphysical project to which the Thesis is intended as a contribution: Obviously, if we use the sentence Snow is white to express the thought that snow is white, no theory that says it means something else can be correct; just as obviously, nothing of metaphysical interest follows. So a conception of use t to serve the purposes the Thesis is intended to serve can not help itself to
I have discussed some of McDowells reections on perception in Heck (2000). Dummett concedes this point in his reply to In Defense of Modesty (Dummett, 1987, p. 257). 6 I should also argue that some of McDowells discussions of perception run epistemological issues together with metaphysical ones (as do those of some of other recent defenders of direct realist views about perception). On this point, see Peacocke (1992, p. 238, n. 22).
5 4

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

facts about the meanings of sentences or other expressions. (It is another question, of course, what an appropriate notion of use might look likea question to which I shall turn in the next section.) Now, McDowell could (and, I think, would) concede that the Use-Meaning Thesis is intended to be a metaphysical principle in this sense. His | view, as I 534 understand it, is that the impression that there is a substantial question about what it is for words to mean what they do depends somehow upon the sense-datum conception of understanding. And so, he might say, I have done nothing to rebut his argument, since I have not provided any alternative way of motivating this problem, that is, of explaining why one ought to think there is a problem about what it is for words to mean what they do. But I nd it difcult to understand how liberation from the sense-datum conception is supposed to dissolve this problem and even harder to know how to evaluate a claim of this kind. And, although I believe there is more that could be said to motivate the metaphysical project, one part of me wants simply to say that it is obvious that there is a real problem about the nature of meaning and that, if one wants to insist upon assigning the burden of proof here, it is far from clear that it should not go to McDowell. But this response would miss part of McDowells point, for he does offer positive reasons to think that the question what it is for expressions to mean what they do need not be taken seriously. What I shall claim in response is that his arguments rest upon a confusion about the nature of homophonic theories of truththeories whose meta-languages include their object-languagesand upon a failure to appreciate the distinction between two different, but intimately related, sorts of questions: Semantic questions, such as what the reference of a particular expression is, or what logical form a particular sort of sentence should be taken to have; and meta-semantic questions, such as what it is for a particular expression to have the reference that it does.7 It would be wrong to say that McDowell does not recognize this distinction at all, or that he rejects meta-semantic questions as unintelligible (though he does sometimes give that impression). In fact, he does offer an answer to the question what it is for a sentence to mean what it does, namely: A sentence S means that p iff the assumption that S means that p should t into a wider context, in which the speakers behavior in general, including their linguistic behavior and their non-linguistic behavior, under suitable descriptions, can be made sufciently intelligible in light of propositional attitudes. . . whose ascription to them is sufAs I shall suggest below, part of the problem here is an ambiguity in the term theory of meaning. I shall use the term theory of truth to refer to semantic theories and the term theory of meaning to refer to meta-semantic theories.
7

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

ciently intelligible in light of their behavior, again, and of the facts which impinge on them. (McDowell, 1998e, pp. 445) McDowells answer to the meta-semantic question is thus a thin one: To say | that 535 snow is white means that snow is white is just to say that taking it to mean that makes speakers sufciently intelligible. Once we recognize this point, though, it becomes somewhat hard to understand how McDowell can make the sorts of epistemological objections he does to Dummetts position. Some of McDowells writings make it seem that he takes Dummetts claim that facts about what speakers mean by a given sentence are constituted by facts about how they use it to be undermined by the observation that our apprehension of what speakers mean is in no way dependent upon our apprehension of how they use it. But if that were his objection, it would apply equally to his own position: My knowing that some stranger means that snow is white when he says snow is white does not seem to depend upon my knowing that he would be made sufciently intelligible by this hypothesis. The objection McDowell means to raise therefore must be sought elsewhere. McDowell is impressed by a certain feature of homophonic theories of truth, and he suggests that attention to this feature of them, together with recognition that homophonic theories are as good as any, will encourage the proper attitude towards the ambitions of the theory of meaning.8 The thought here is that a theory of truth that reports such facts as Snow is white is true if, and only if, snow is white does not invite perplexity about what it might be for it to be correct: it wears its correctness on its face, as it were (McDowell, 1998c, p. 69). For it is obviously a necessary condition on the correctness of a homophonic theory that it prove homophonic T-sentences, that is, theorems like 2. And if, as one might well think,9 the theorys meeting this condition were not only necessary but also sufcient for its correctness, there would be no deep meta-semantic question to be raised about it: the theory will be correct if, and only if, it proves all homophonic T-sentences; whether it does so is a formal matter, something that can be decided simply by looking at its syntax.10 But if the correctness of the theory is, in that sense, obvious, it is hard to see why the question what it is for the theory to assign the right
The appropriate attitude is what McDowell calls a modest one, and what is at issue here is, in a sense, whether a theory of truth need only be modestthat is, need only state what meanings the speaker takes expressions of her language to haveor must be full-bloodedthat is, must also explain what it is for her words to mean what they do. The issue I am discussing here has been discussed by Dummett on a number of occasions: see Dummett (1993e, 1987, 1991, ch. 5). While I do not really disagree with his diagnosis, I am not sure it goes deep enough; it has the additional disadvantage that it is given from deep inside his own position. 9 There are complications here, but they do not affect the points being made. See note 15. 10 This kind of point is familiar from discussions of Tarksis theory of truth. See, for example,
8

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

references to expressions should be taken seriously: The sort of meta-semantic question McDowell nds intelligible therefore has a trivial answer. I do not want to object to McDowells claim that the correctness of a homophonic theory of truth can, in a certain sense, be determined by reection. It is worth noting, however, that, in establishing the truth of 2 by such reection, we draw upon information that is not contained in it. One must realize that the sentence named on the lefthand side is the same as the sentence used on the right and not just that it is the same sentence, in a syntactic sense, but that it has the same meaning. Nothing in the T-sentence tells one that: One could have a perfectly good understanding | of the sentence itself, and yet not realize that the same sentence is 536 both used and mentioned. One can build this information into the T-sentence this way: The sentence on the right-hand side of this very biconditional is true, in the very language I am now speaking, if, and only if, snow is white. And it is at least arguable that the truth of (2) will be completely obvious to anyone who understands it, although it is in fact equivalent to the original T-sentence 2, as that sentence is intended to be understood. Note, however, that the fact reported by 2, and so by 2, is contingent.11 It is obviously a contingent fact about my idiolect that the sentence Snow is white is true in it if, and only if, snow is white.12 It is this sort of fact that a theory of truth will report and partially explain, by deriving it from others about the meanings of sub-sentential expressions and grammatical constructions; and it is this contingent fact that 2 and 2 report. This would once have been confusing: how can it be both that 2 is contingent and that its correctness can be determined simply by reection? But, thanks to Kripke, this is now a familiar sort of puzzle. Consider the sentence I am here. Just by reecting upon it, one can easily convince oneself that any utterance of it must be true. And yet the fact such an utterance would report would (typically) be contingent. That is to say, any particular utterance of the sentence would provide one with an example of a contingent a priori truth. Whether one really wants to say that T-sentences like 2 and 2 are contingent a priori (perhaps they are not really a priori at all), they are cut from the same cloth: They are sentences whose truth can be determined purely by reection, and yet the facts they report are contingent ones. McDowells view, as said above, is that, once we see that there is no reason not
Etchemendy (1988) and Soames (1984). 11 The point being made here has its origin in Tarskis distinction between formal and material adequacy and in his separation of the mathematical from the empirical aspects of semantics. For more on this, see Heck (1997). Compare Dummett (1991, pp. 6971). 12 I might have used the word white to mean black, and then Snow is white would have been true if, and only if, snow were blacka fact I could then have reported, of course, by uttering 2.

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

to state a theory of truth as a homophonic theory, and see also that the correctness of a homophonic theory is something that is obvious to anyone who understands it, we ought to realize that there can be no serious question about what it is for it to be correct. But this argument nishes with a non-sequitur. If someone asks Why am I here? it is no answer to tell her that here conventionally refers to the place of utterance.13 She knew that: What she wanted to know why she was in the place she happened to be at that time, which she picked out via the word here. And if someone asks Why is Snow is white true if and only if snow is white? it is no answer to tell him that the word true is conventionally used so as to sustain disquotation. He knew that: What he wanted to know was why the truth of the sentence Snow is white stands or falls with the whiteness of snowrather than, say, with the greenness of grass. These questions should not be confused with the historical question | how the 537 sentence snow is white came to mean what it does. That question might be answered etymologically, and, of course, the answer is not likely to be of great philosophical interest. Rather, the question being asked is what it is about an utterance of this sentence that makes it mean anything at all, and what it is that makes it mean the very thing it does. The question, that is to say, is what the relevant difference is between snow is white and blurg is white; and between snow is white and grass is green.14 These questions are not themselves philosophical ones: They presumably have answers that will one day be provided by some (perhaps now non-existent) branch of empirical linguistics. But its not just that we dont know the answers to these questions. The problem is much worse than that: It is that we do not even have any very good idea what the answers to them might look like (which is not to say that no one has any ideas about the matter). We do not really know what properties of the sentence snow is white are semantically signicant. And in so far as we do not know that, we do not know how its semantically signicant properties, whatever they may be, conspire to x its meaning. This general ignorance is the source of philosophical problems about the nature of linguistic meaning: To ask what properties of an expression contribute to determining its meaning, and how they determine it, is precisely to ask the meta-semanticthat is, the metaphysical question what it is for an expression to mean what it does. The Use-Meaning Thesis, as I understand it, amounts to a rough outline of an answer. When we raise semantic questions, such as what logical form the sentence
I owe this example to Vann McGee. It inspired a much cleaner treatment of this entire issue. Note that one might put this point by saying that all that has actually been explained is why an utterance of the sentence I am here is true, not why I am here. In so far as that is right, this sort of point is close to those made by Dummett in his discussions of this issue (1993e, pp. 812). 14 For some further discussion of these matters, see Heck (2004).
13

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

Snow is white should be taken to have, we need not simultaneously raise metasemantic questions, such as what it is for the sentence to have the truth-condition it does. Even if we operate under the assumption that the meta-language extends the objectlanguage, so that the T-sentences the theory proves can be seen to be correct merely by reection, the problem of constructing a theory of truth (for a natural language) that actually has all homophonic T-sentences as theorems is non-trivial. And although no serious semantic theory will be close to homophonic (indexicality already makes this impossible),15 whatever vocabulary is common to the object- and meta-languages is typically assumed to have the same meaning in each. (That will simplify the task of evaluating the theorys correctness, even if it does not reduce it, in all cases, to a problem in proof-theory.) But that does not imply that meta-semantic questions are not there to be raised: and it does not imply that it is legitimate to appeal to the assumption that the meta-language extends the object-language in the context of meta-semantic investigations. To appeal to that assumption, and so to answer the question | what it is for Snow is white to be 538 true if and only if snow is white by alluding to disquotation, is to fail to answer the question being asked. It is instructive here to consider Donald Davidsons early writings on the theory of meaning. Discussing homophonic theories of truth, Davidson writes in Truth and Meaning that the trouble is to get a theory that comes close to working; anyone can tell whether it is right (Davidson, 1984c, p. 25). And yet, he also insists that a theory of truth is supposed to be an empirical theory about the semantic properties of the expressions in a particular language (Davidson, 1984c, p. 245, 27).16 My point is not that Davidson is contradicting himself: we have seen that these views are compatibleif, by an empirical theory, we mean one that states contingent facts about its subject matter. It is, rather, that Davidsons expressing both of these views is an indication that he is interested in two different projects. The rst is the semantic project of actually developing a theory of truth for a natural language, that is, a theory sufcient to yield theorems stating the semantic
It is one of the ironies of the subject that there is no agreement at all on how mass terms should be treated, that is, how a T-sentence for Snow is white and other sentences containing snow might actually be derived. Actual semantic theories for such sentences tend to make heavy use of concepts from Boolean algebra, and the T-sentences they deliver are not even close to homophonic. One can, of course, add axioms to force the theories to be homophonic, but doing so does not increase the interest of the theories. Moreover, semantics is interested in saying what different readings might be available for a given sentence: It is important that the sentence I almost had my wallet stolen admits of three different readings; to capture the difference, the theory will have to be non-homophonic. See Higginbotham (1995), from which I borrow the example just used, which originates with Chomsky. 16 See also (Davidson, 1984b, pp. 1345). Davidson might be taken to mean by empirical a posteriori, in which case there really is a contradictionunless the claims are taken to concern different projects, as below.
15

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

properties of all expressions of English (and to systematize that collection of facts by deriving those concerning complex expressions from those about their simpler parts). The second is the meta-semantic project of answering the question what it is for English expressions to mean what they do. The two projects are not always clearly distinguished in Davidsons writings, since, for him, they are intimately related. Truth and Meaning is, I think it fair to say, most famous for Davidsons having proposed there that a theory of meaning may take the form of a Tarskian theory of truth. But something that is, for the most part, only implicit in the paper is more signicant. In 1967, semantics was still in its infancy, and its inuence on philosophy of language was limited: Quines discussions of the nature of meaning, for example, are quite independent of anything we would recognize as semantics. And yet, Davidsons whole approach, made more explicit in later writings, rests upon the thought that questions about the nature of meaning should be approached by inquiring into the form that might be taken by a complete theory of meaningthat is, a semantic theoryfor a natural language.17 Since a theory of truth, as Davidson conceives it, is a theory that proves sentences giving the meanings of all signicant expressions of a given language, the metaphysical question what it is for an expression to mean what it does may be re-cast as: What is it for a theory of truth to be correct? Asking the question in this way forces us to base our answers to meta-semantic questions upon a conception of what semantic facts are that is clear enough, at least, for the purpose of developing a compositional theory of meaning for a complete language;18 and it enables us to answer the question what it is for Snow is white to be true if and only if snow is white, by addressing the | more specic questions what it is for its seman- 539 tically signicant parts to mean what they do and for the modes of composition to have the signicance they do. To put the point crisply: Davidsons proposal is that meta-semantics should be meta with respect to semantics. And, in the context of the meta-semantic projectwhich, as argued above, is the only one within which the Use-Meaning Thesis has any placeDavidson commits himself to a version of it. His view is that the correctness of a theory of truth is determined by the data available to a radical interpreter prior to, and indepenThe claim is implicit in Davidsons suggestion that the correctness of the theory can be tested using something like radical translation (Davidson, 1984c, p. 27). It is made somewhat more explicit in Radical Interpretation (Davidson, 1984b), where Davidson argues that a theory of truth can serve as a theory of interpretation.I owe my own appreciation of this point to Higginbotham (1988). 18 It is this issue that Davidsons other proposal addresses, for it embodies the suggestion that semantic facts are facts about the references of expressions, in the broad sense in which Frege used that term. It also reects a commitment to Freges context principle, that is, to the priority of sentencemeaning over word-meaning.
17

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

dently of, any understanding she might have (or come to have) of the language the theory concernsthat is, that a theory of truths correctness is determined by how the object-language is used, where the notion of use is explained in terms of the data available to a radical interpreter. Davidson explicitly denies that these data include facts about what particular expressions mean. The reason that he restricts his notion of use in this way, and thereby rejects meaning-laden notions of use, is just the one mentioned earlier: To characterize usethat is, the datain terms of the meanings of sentences would be to trivialize the claim that the semantic properties of expressions are determined by how they are used and, with it, the meta-semantic project of Radical Interpretation (Davidson, 1984b, pp. 1345). If one runs the semantic and meta-semantic projects together in ones mind, it will indeed seem perplexing how one could be interested in homophonic theories of truth, as Davidson is, and yet think, as he also does, that there is a real question about what it is for a theory of truth to be correct, that is, that there is any problem to which the Use-Meaning Thesis might be the solution. But once the distinction between semantics and meta-semantics has been drawn, McDowells assumption that homophonic theories have any special relevance to meta-semantics can be seen to rest upon a failure to appreciate it. Only if we assume that the meta-language extends the object-language can we reectively convince ourselves of the truth of homophonic T-sentencesonly then will consideration of such theories seem to cast doubt upon the depth of meta-semantic questions. But one can not appeal to that assumption in answering the meta-semantic questions to which the Use-Meaning Thesis provides an outline of an answer. (That is just to say that the original, at-footed response to McDowell, that the Use-Meaning Thesis is a metaphysical principle, and not an epistemological one, stands.) We are now in a position to return to the question on what grounds Dummett rejects meaning-laden notions of use. Dummett is primarily interested in questions about understanding, that is, in questions about linguistic and, specically, semantic competence. And his proposal is that a speakers semantic competence should be explained in terms of her knowing a theory of truth for her language, that is, in terms of knowledge | she has about her languages semantic properties. Dummett 540 does not, however, mean to be ignoring questions about the nature of meaning: his view, rather, is that . . . the key to an account of language. . . is the explanation of an individual speakers mastery of his language (Dummett, 1993d, p. 99). That is to say, he construes the question whether a theory of truth for a given speakers language is correct as the question whether she knows itwhether it states what she knows about the semantic properties of expressions of her language, in virtue of her knowing which she is a competent speaker.19
19

It is, as he puts it, not a description from the outside of the practice of using the language,

10

Richard G. Heck, Jr. 2

Use and Meaning Use as Meaning: Metaphysical, Not Epistemological

The suggestion that speakers know theories of truth for their languages raises some very hard problems.20 It is far from clear what it might mean to ascribe knowledge of such a theory to a speaker, since ordinary speakers obviously do not consciously know theories of truth for their languages.21 Now, the usual alternative is to suppose that such knowledge is tacit or implicit. But, even if we assume objections to the very intelligibility of the notion of tacit knowledge to have been answered, and even if we assume that we understand in what it might consist that a speaker has tacit knowledge of the semantic structure, or logical form, of sentences,22 the question remains what it is for a speaker tacitly to know a particular theory of truth. For theories of truth may differ, even if they agree about logical form: They may have different axioms, leading the theories to output different T-sentences as theorems. The theories might output T-sentences that differ in truthvalue, and it has not even been said, at this point, what it might be for one, rather than another, of these theories to be that tacitly known by a speaker.23 Thus, the meta-semantic problems do not vanish once we assume that speakers know theories of truth for their languages; rather, they are transformed into questions about what it is for a speaker to know a particular such theory. Dummetts answer to such questions is offered in passages like the following: [T]he philosophical task of explaining in what a mastery of a language consists is not completed when we have set out the theory of meaning for the language. . . . [W]e have to go on to give an account of what it is to have such knowledge. This account can only be given in terms of the practical ability which the speaker displays in using sentences of the language. . . . (Dummett, 1993d, p. 101) Now, it is tempting to interpret Dummett as claiming that a speakers knowing a particular theory of truth consists in her having certain linguistic capacities. And,
but. . . an object of knowledge on the part of speakers (Dummett, 1993d, p. 100). 20 These problems do not arise for Davidson. The problem facing the radical interpreter is to discover a theory her knowing which will make her a competent speaker. Davidson is free to suppose that the interpreter already understands a language in which the theory of meaning can be formulated. But for that very reason, the project of radical interpretation does not address questions of competence. This, indeed, is Dummetts most fundamental criticism of Davidsons program (Dummett, 1993e, p. 6). 21 In a certain sense, it would not matter if they did. The knowledge that underlies competence can not be conscious, explicit knowledge: Even if Anglophone semanticists should one day give a complete theory of truth for English, it would not be their explicit knowledge of that theory that explained their competence (see Dummett, 1993e, pp. 21ff). 22 For discussion of this, see Evans (1985); Davies (1987); and Wright (1993). 23 Of course, it is also possible for the theories to have different axioms, but for the T-sentences output to agree in truth-value, even necessarily. This problem, the Foster problem, is part and parcel of the problem I am discussing. For my own preferred resolution of it, see Heck (2005).

11

Richard G. Heck, Jr. 3

Use and Meaning Language, Thought, and Use

in What is a Theory of Meaning?, he writes that a theory of truth is a theoretical representation of a practical ability, the implication being that a speakers knowing the theorythat is, her understanding her languagesimply consists in her being able to speak it (Dummett, 1993e, p. 21). On this view, | the structure the 541 theory assigns to sentences has little purpose other than to articulate this complex ability into component sub-abilities, possession of which constitutes the speakers understanding of the primitive expressions; these abilities then interact, in a way captured by the deductive structure of the theory, to produce other abilities, constituting the speakers understanding of sentences (and other complex expressions). Talk of a speakers knowing the theory then becomes idiomatic: Speakers know the theory simply in the sense that it articulates an ability they possess; someone who did in fact know it would be able to speak the language. If a theory of truth amounts to an articulation of a speakers complex ability to speak her language into simpler abilities that jointly constitute it, it must surely follow that the correctness of the theory lies in its correctly characterizing this ability, in a correspondence between the references the theory assigns to the primitive expressions and aspects of the practice itself. That is to say, for Dummett, a description of the use of an expression will be an account of the ability a speaker must possess to understand it. How exactly these abilities are to be characterized remains open, but, as with Davidson, what is motivating Dummetts selection of a notion of use are meta-semantic concerns. This may be obscured by the fact that he argues for the Thesis by arguing that we have an obligation to say what it is for a speaker to know such a theory. But, as said above, on Dummetts view, to state the theory known by competent speakers is to say what those expressions mean; to say in what it consists that a particular such theory is that known by competent speakers is to say in what it consists that they mean what they do. Since, on Dummetts view, to know such a theory is be able to engage in the practice of speaking the language, the description of the practicethat is, speakers use of the languageserves to ground an answer to the question what it is for expressions to mean what they do. To describe use in terms of the meanings of expressions would therefore be to render the Use-Meaning Thesis trivial, and that is why Dummett rejects meaning-laden notions of use.

Language, Thought, and Use

McDowells challenge was to explain how the Use-Meaning Thesis can be both substantial and plausible. To this point, my task has been primarily defensive: I have argued only that a rejection of meaning-laden notions of use ows from an interest in meta-semantic questions and need not be grounded upon any particu-

12

Richard G. Heck, Jr. 3

Use and Meaning Language, Thought, and Use

lar view about the epistemology of understanding. But that will not much matter unless there are other notions | of use available, besides the Quinean one and the 542 meaning-laden one McDowell offers us, and I have yet to argue that there are any other alternatives. It is to that task that I now turn. One might have hoped that an alternative notion of use could be extracted from the position with which we left Dummett at the end of the last section, from the idea that to understand a language is to possess a practical capacity to speak it. But McDowells most important objection to Dummett is that no notion of use so obtainable will be different enough from Quines. As McDowell sees it, and I would agree, the problem with the Quinean notion is that, if we conceive of the use of language behavioristically, we will be without any explanation of its rationality, without any way of recognizing the fact that the use of language is a rational activity on the part of rational agents.24 And, or so he suggests, a similar problem will afict Dummetts position, that his notion of use too can be [no] more than a mere description of outward behaviour, with the mental. . . aspect of language use left out of account. McDowells view, of course, is that the only way of registering the role of mind in our use of language is by describing use in terms of the contents of speech acts (McDowell, 1998c, p. 65).25 But, or so I shall argue, once we have a proper understanding of the signicance of McDowells objection, the way to a better account will be clear. Something like McDowells objection is familiar from the writings of Noam Chomsky, who has argued repeatedly that Dummetts view, like any that identies linguistic competence with a practical ability to speak a language, fails to make sufcient room for speakers knowledge of their language (see, for example, Chomsky, 1980, ch. 2). As said earlier, if we conceive understanding as a practical ability, talk of a speakers knowing a theory of meaning becomes a faon de parler, and there are a variety of reasons to think we have to take speakers linguistic knowledge more seriously than that. Dummett himself has not been insensitive to such concerns.26 In some of his discussions, he expresses the worry that conceiving of understanding as a practical capacity would prohibit us from giving due recognition to the fact that speakers typically know what they mean. Thus, he notes that, on Grices explanation of
Dummett himself emphasizes this fact: See, for example, e.g. (Dummett, 1993d, p. 104). See also McDowell (1998a). Of course, no one is suggesting that this objection would impress Quine. 26 Just a few years after the publication of What Is a Theory of Meaning? the objection I am about to rehearse in fact leads Dummett to reject the view that to understand a language is to possess a practical capacity to speak it. It seems fair to say, however, that Dummett has never settled upon an alternative view. Even the appeal to implicit knowledge is deemed unhelpful by the publication of Logical Basis.
25 24

13

Richard G. Heck, Jr. 3

Use and Meaning Language, Thought, and Use

conversational implicature, speakers intentions play a crucial role; and speakers must know what they are saying, and what is being said to them, if Grices account is to be correct, since only what a speaker consciously knows about the meanings of her words can play a role in the formation of her linguistic intentions. We do not recognize the same implicatures in the speech of those whose knowledge of English is poorbecause they have no knowledge of the subtleties of meaning necessary for the formation of the requisite intentions (see Dummett, 1991, pp. 912).27 I do not want to quarrel with Dummetts, or Chomskys, claim that we should 543 take speakers knowledge of their language more seriously than Dummett was once inclined to take it. But I do not think this sort of observation goes to the root of the problem. The problem, as I understand it, does not concern any specically linguistic knowledge we might have, either tacit or conscious, but general facts about how our use of language is integrated with our conscious mental life.28 It is not entirely clear what is meant by saying that understanding is a practical capacity, but the essential feature of the view, as Dummett conceives it, would appear to be that what a practical capacity is a capacity to do can be fully characterized in terms independent of the agents conscious psychological states. Consider, for example, the ability to swiman example of a practical capacity, if anything is. To swim is, roughly, to locomote in water, so only something capable of locomotiona creature, let us sayis capable of swimming. But swimming is not necessarily rational action: Fish can swim, and saying that they can does not commit us to viewing them as conscious, let alone rational, creatures. That is not to say that the ability to swim is not to be explained, empirically speaking, in terms of information-processing states within a creatures mind: But there is no reason such states need be accessible to consciousness. If the use of language were an exercise of a practical capacity in this sense, what it was to use language would be fully characterizable in terms independent of a speakers conscious psychological states. Now, it is true that, in many respects, our ability to use language depends upon sub-personal information-processing: For example, if contemporary linguistic theory is on anything like the right track, our perception of syntactic structure is due to the functioning of complex, and largely innate, systems of whose workings we ordinarily have no conscious knowledge. But our use of language is not all like that.29 There is a strong intuition that only
A similar point can be made about speakers with only partial knowledge of meaning: If a rheumatologist were to report that he had arthritis in his thigh, one would presumably wonder what else he actually meant to convey; but someone with only partial understanding of the term should not be so taken. 28 For further discussion of this matter, see Heck (2005). 29 It is for this reason that I think Chomskys observations, valuable as they are, do not really speak
27

14

Richard G. Heck, Jr. 3

Use and Meaning Language, Thought, and Use

rational agents are capable of using language, that one is not using language unless what one says is connected, in the right kind of way, with what one thinks. I do not mean here that one ought only to say what one believes: I mean something much more fundamental.30 It is one thing to suppose us capable of having thoughts we are unable to express in language; but the converse, that we should be capable of expressing a content without also being able to have thoughts with that same content, is absurd.31 If our linguistic abilities were really comparable to the ability to swim, then a sentences having a particular content would be but a matter of our using it in a particular way, where what it was to use it in that way would be something that did not require us even to be able to have any thoughts with the content in question. But our use of language isand if it is to count as use of | language at all, must beintegrated with our conscious mental life, at least 544 in the minimal sense that the contents of our utterances must be able to gure in our thoughts. And the fatal problem with the conception of understanding as a practical ability is that it utterly precludes any account of this integration.32 If we are to secure languages integration with thought, we must recognize it from the very outset: As James Higginbotham puts a similar point, even if our understanding of our language consists in our capacity to use it, the relevant capacity must involve the capacity to judge that such-and-such under particular conditions (1994, p. xxx). This observation might well be used to motivate a quite different conception of understanding, and so a correspondingly different conception of use. On this view, our understanding of language depends upon a prior, and independent, capacity for thought, in the sense that a prior capacity to have thoughts with particular contents is required if one is to be able to use expressions with those same contents; it is this prior capacity that must be invoked if we are to explain what it is for those expressions to have those contents. One might, indeed, be attracted to such a position for other reasons. The simplest is just that beings without language seem to be capable of having certain sorts
to the fundamental problem: True though it may be that our use of language has to be conceived as resting upon our possession of tacit knowledge, this observation does not address the rationality of language-use, for a similar point might be made about such abilities as the ability to swim. Indeed, Chomsky himself makes just that point (1988, pp. 1112). 30 I am indebted here to Dummetts discussion in Logical Basis (1991, pp. 901), although it is again bound up with questions about whether we know what we mean. 31 I am, of course, ignoring cases in which speakers utter sentences they do not understand. And one should not really say, at this point, that the thoughts need to have the same content: On some views, the contents of thoughts are of a different sort from the contents of sentences, and so the right thing to say will be that we must be able to have thoughts with appropriately related contents. See Heck (1995). 32 Someone sympathetic with this view might try to answer this objection by making use of the sort of machinery discussed in Section 4. But once that has been done, the claim that understanding is a purely practical ability has been abandoned.

15

Richard G. Heck, Jr. 3

Use and Meaning Language, Thought, and Use

of thoughts: Pre-verbal children, for example, seem to be able to have thoughts about their mothers, about the colors of objects, and so forth.33 And if that is right, it would, to say the least, be surprising if the childs developing understanding of her language did not, in some way, exploit her prior capacity for such thoughts. Moreover, there must be a connection between these thoughts and the sentences that come to express themthat, after all, is the point of the remarks about integration aboveand one way to secure it will be to insist that what it is for the child to mean red by red has to be explained in terms of her prior grasp of that very concept, in terms of a connection between her beliefs (and other attitudes) involving this concept and her use of sentences containing the word.34 It is important to recognize that assuming a prior grasp of content does not dispose of the meta-semantic question what it is for an expression to mean what it does. The assumption that Kurt can entertain the thought that snow is white does not, of itself, yield any answer to the question what it is for him to express that thought by means of some form of words.35 Indeed, what Dummett takes to be the fatal objection to such views is precisely that they offer no answer to this questionnone more promising, in any event, than is contained in the thought that we manage, somehow or other, to associate the content with the linguistic expression (Dummett, 1993d, pp. 979). But Dummett wrongly supposes that such views can give no answer to this 545 question because he overlooks the fact that it is possible both to accept the priority of thought over language and to embrace the Use-Meaning Thesis: To do so, one need only insist that use be characterized in terms of the contents of mental states; we might call such a notion of use a Gricean one. Such a specication of the Thesis will not enable one to answer the question what content is, in general: But, on this view, the philosophy of language is not where that question ought to be answered, and the Thesis can yet play a role in an explanation of what it is for a
A common move here is to deny that pre-verbal children have anything properly called thoughts at all. Dummett himself makes this move (1993c, pp. 1489). The difculty is that, even if this is right, it is hard to see why a capacity for proto-thought, if that is what one wants to call it, should not still be invoked in explaining our grasp of our most primitive concepts. Is one so sure that proto-thoughts are not just non-conceptual contents by another name? 34 One version of this view has it that a theory of truth for English is explicitly represented in our minds in the language of thought. See Segal (1994) and (1995, ch. 1). But the means of representation is not crucial here. 35 Even if we grant that Kurt knows that Snow is white is true if, and only if, snow is white, the question remains how this knowledge is to be deployed, how Kurts having it gives rise to the linguistic abilities that manifest (even if they do not constitute) his competence. In principle, Kurt could have such knowledge and have no idea what the sentence means. More therefore needs to be said before even the view mentioned in the preceding note will be of any help here. For some of it, see Higginbotham (1992).
33

16

Richard G. Heck, Jr. 3

Use and Meaning Language, Thought, and Use

linguistic expression to mean what it does.36 The very observation that shows that the meta-semantic questions do not vanish the moment we assume the priority of thought also shows that characterizing use in terms of the contents of mental states does not trivialize the Use-Meaning Thesis. Adopting such a characterization of use does affect the metaphysical force of the Thesisin particular, it no longer promises a solution to the problem of intentionalitybut it does not drain it of force altogether. There are various ways a Gricean conception of use might be developed. The general thought, to put the point in Davidsonian language, is that the data available to the radical interpreter should be characterized, not in terms of the external conditions under which speakers make various utterances, but in terms of the psychological conditions (in addition, perhaps, to external ones) under which they make them. One option would be to pursue some form of Grices programhence my calling the notion of use in play here a Gricean onearguing that Kurts meaning that it was raining consisted in his uttering It is raining with the intention that his audience should come to believe he believed that it was raining, by means of their recognizing his intention that they should, and so on and so forth. Another would be to pursue a proposal of Ian Rumtts, which makes essential use of the thought that speakers know the truth-conditions of utterances, in the sense of knowing Tsentences for them (Rumtt, 1995). There are other possibilities, to be sure. But the question how such a position is best developed is not the one I wish to pursue here: The issue I want to discuss is a more abstract one that, I think, troubles Dummett in particular. One might worry that this way of accommodating the rationality of our use of language will commit us to an implausibly strong version of the claim that thought is prior to language. Let us grant that there are many sorts of thoughts pre-verbal children are capable of having and, moreover, that any explanation of what they mean by their most basic utterances will have to make reference to this prior capacity. It is far from obvious that anything like this is correct in general: Surely it is just false that, for any sentence a speaker might come to understand, she must have a prior | capacity to entertain the thought it expresses. For example, 546 it is far from clear that speakers acquire a capacity to make reference to arbitrarily distant regions of time before they acquire a capacity to use the past-tense. A similar view is yet more plausible in the case of all but the most basic mathematical notions.37 And, even if neither of these examples is compelling, it would be unfortunate if answering McDowells objection required us to place a bet that there are
I thus think that Dummett misunderstands Grices intent. Dummett (1993a, pp. 1713). Note that the question in play at this point is the empirical one whether we in fact have a prior grasp of certain contents. The mere fact that one might grasp the contents of certain expressions without having a means for expressing those contents does not decide the question how our actual understanding of those expressions is to be explained. Compare Peacocke (1997).
37 36

17

Richard G. Heck, Jr. 3

Use and Meaning Language, Thought, and Use

no better ones. If there are sentences our understanding of which is not to be explained in terms of a prior capacity for thoughts with the same content, some notion of use other than the Gricean one will be needed. But the relevant notion of use need not be entirely meaning-free. It would, of course, undermine the purpose for which the notion of use is wanted if, in characterizing the use we make of past-tense sentences (to take them as an example) reference were made to the meanings of sentences in the past tense (which would trivialize the Use-Meaning Thesis) or to the contents of thoughts about the past (which would return us to a Gricean notion of use). But it would not be similarly illegitimate to make reference to the meanings of the present-tense transforms of those same sentences: The explanation of what it is for the past-tense sentences to mean what they do will have to advert to the meanings of the present-tense sentences embedded in them, since an understanding of past-tense sentences is surely parasitic upon an understanding of their presenttense transforms. And note too that we need have no qualms about making use of psychological notions, like belief, intention, or what have you, so long we do not assume that the speaker is antecedently able to have thoughts about the past. Such a notion of usewhich I shall call a Dummettian notionwill thus not be vulnerable to McDowells objections to the Quinean notion of use. We will be able to recognize the integration of a speakers understanding of sentences in the past-tense with her capacity for thoughts about the past, so long as the description of use makes reference, for example, to a requirement that speakers beliefs reect their assertions, and those of others, in an appropriate way. One way to develop the Dummettian notion of use might be to take as a model Peacockes use of psychological notions in stating what he calls possession conditions: It is essential to a possession condition that it make reference to the agents beliefs, for example; but it is just as essential that it not presuppose a capacity for beliefs whose contents contain the concept the possession condition concerns. The same sorts of moves Peacocke uses to resolve the threat of circularity in that case (Peacocke, 1992, pp. 610) are also available for framing a Dummettian notion of use.38 Another option might be to formulate the requirement that the use of lan38 McDowell thinks there are serious problems with Peacockes non-circularity condition, a commitment to which amounts, in effect, to a rejection of content-laden notions of use. But I think McDowells objections can be answered. His discussion overlooks the crucial role played by the requirement that a determination theory be provided for any putative possession condition: It is this, not some need for possession conditions to be given within the space of reasons, that motivates rejection of the neuro-physiological condition McDowell considers in Mind and World (1996, pp. 1678) It is because it is impossible to provide such a determination theoryshort a type-type identity theory for concept possession, which is not even worth discussingthat such an account would have nothing to say about what someone thinks when she thinks that something is red. It is the determination theory that species the content of the concept (what its semantic value is), not the

18

Richard G. Heck, Jr. 3

Use and Meaning Language, Thought, and Use

guage be integrated with thought as a general principle that does not enter into the characterization of the use of particular expressionsthat is, as a general constraint on what it is for an expression | to mean anything at all, rather than as something 547 that enters into specic accounts of what it is for expressions to mean what they do. But let me not pursue this issue further here. The overall shape of one sort of meta-semantic project based upon the UseMeaning Thesis may thus be understood as follows: For the most fundamental parts of our language, where we have a prior capacity for thoughts with the contents in question, what it is for sentences to mean what they do will be explained in terms of some version of the Gricean notion of use; for other parts, if there are any, in terms of the Dummettian notion. To decide what notion of use to employ in specifying the Use-Meaning Thesis, we must thus decide whether, and where, thought is prior to language. There is no reason such a decision needs to be made once and for all: Although there are extreme views that merit consideration, the truth may well lie somewhere else.39 The availability of the position just described seems to me enough to undermine McDowells claim that, once we recognize that the view that understanding is a practical capacity is unable to account for the rationality of our use of language, our only option is to describe use in terms of the contents expressed by utterances. The Gricean conception of use is content-laden, to be sure, but it does not help itself to any notion of linguistic contentthat is, it is not meaning-ladenand the position is designed precisely to allow us to recognize the integration of language with thought. On the other hand, however, even if this point is accepted, McDowell might still be said to have offered a compelling argument against Dummett. For Dummett is not prepared to employ a Gricean notion of use anywhere, since he rejects the priority of thought, arguing that it is at best useless to appeal to a speakers prior possession of particular concepts in attempting to explain what it is for her words to express them. It does not follow, however, that any notion of use Dummett could accept would have to be content-free, or even meaning-free: In particular, in explaining what it is for past-tense sentences to mean what they do, he could deploy a notion of use that was not meaning-free, for the reasons discussed above. In general, whenever a speakers understanding of one (sort of) sentence depends upon a prior understandpossession condition on its own. 39 One might wonder whether there is not another notion of use to be explored, one that would allow use to be characterized in terms of the general notion of linguistic meaning, but would not appeal to any particular facts about linguistic meaningmuch as the Dummettian notion appeals to psychological notions, but not to our capacity for particular thoughts. Such a notion of use is indeed available, in principle, but one might worry that, by taking the notion of linguistic meaning for granted, it would enable us only to answer specic questions about what it is for words to mean what they do, and not to answer general questions about what it is for words to mean anything at all. But the matter is complex, and only a developed proposal would allow us to resolve it.

19

Richard G. Heck, Jr. 3

Use and Meaning Language, Thought, and Use

ing of another, it will be legitimate for Dummett to advert both to what she means by that other sentence and to her capacity for thoughts whose content it serves to express.40 But that is all just a sideshow. The real issue concerns what goes on at the most fundamental level. For sentences at that levelsentences our understanding of which does not depend upon a prior understanding of any other sentences the notion of use we employ in explaining what it is for those sentences to mean what they do will have to be both meaning- and | content-free. One might well 548 worry that, at this point, Dummett is forced to turn back to the conception of understanding as a practical ability and face McDowells objection anew. He writes that some [of a speakers knowledge of her language]the deepest and most interesting components[consists] of a complex of acquired practices that together constitute a grasp of content. Dummett does later deny that her understanding consists in the mastery of a purely external practice (Dummett, 1991, pp. 102 3), the thought being that the practice has, so to speak, one foot in her mouth and another in her head. But he does not attain a clear conception of what this amounts to. The reason, I think, is that he fails to distinguish the priority of thought from what we might call the priority of reason. It is one thing to say that thought is prior to language at the level of content: that a prior grasp of the contents of sentences needs to be invoked in explaining our understanding of them. It is entirely another to say that the general capacity for thought and reason is prior to any linguistic capacity. If we accept this latter claim, as it seems to me we must, then we are free to characterize use in terms of psychological notions, like those of belief and intention, and yet refuse to appeal to any prior capacity to entertain thoughts with the very contents expressed by our utterances. In that way, one might recognize Higginbothams pointthat the practice of speaking a language involves a capacity to judgeand yet deny that the practice must be described in terms of an antecedent capacity to entertain the contents we do judge. What Dummett does say about the matter is, I think, also best understood in this light. Dummett says, for instance, that, to understand a sentence, I must know what bearing its truth may have on my actions. In the relevant sense, though, the truth of a sentence will affect my actions only in so far as I believe it to be true: What I need to know, then, is what I would be believing if I were to accept the sentence as true; obviously, that presupposes that my acceptance of its truth will give rise to a belief with an appropriate content.41 But if that is the sort of view
40 Of course, it is here essential that our language have a certain sort of hierarchical structurethat, modulo local holisms, an ordering of the sentences in terms of whether an understanding of the one depends upon an understanding of the other will be a partial ordering. 41 See here Heck (1995), where I argue that this principle is required for different sorts of rea-

20

Richard G. Heck, Jr. Use and Meaning 4 Closing: The Epistemology of Understanding and the Priority of Thought

Dummett envisages, then what he ought to say is that a Dummettian notion of use is to be employed throughout; this will yield a notion of use, at the most fundamental level that, although content-free, is still non-behavioristic and, for that very reason, allows us to recognize the integration of language and thought.42 If we understand Dummett as I have just suggested we should, then his project survives the objections McDowell brings against him. Of course, one might be skeptical that a Dummettian notion of use, at the most fundamental level, can be developed in detail. I have my doubts myself. My point here, though, is that the charge of behaviorismwhich is so often brought against Dummett, with a footnote to McDowells paperscan not be sustained, at least not on the basis of the sorts of general considerations | McDowell develops: No principled reason 549 has emerged that Dummett should be unable to make room for the rationality of language-use.

Closing: The Epistemology of Understanding and the Priority of Thought

As I said at the outset, McDowell suggests that Dummetts rejection of meaningladen notions of use is motivated by a awed epistemology of understanding. I have rejected this reading. But there are, of course, several passages in Dummetts writings on which one might reasonably base such an interpretationpassages where Dummett deploys arguments that certainly do look like epistemological ones in support of the claim that the notion of use should be content-free. In closing, I should like to offer, from our present vantage point, an account of the role these arguments play in Dummetts thought. Illustrating these arguments, McDowell asks us to imagine that Martians speak a language that sounds like English but differs from it semantically: Expressions in Martian do not always mean what their English homophones mean (McDowell, 1998b, p. 244). Dummetts claim about such a case would be that, if meaning were not determined by useconceived in uniformly Dummettian termsit might be impossible for us ever to recognize the semantic differences between the languages. Suppose that the Martians use of their language, so conceived, was in relevant resons, namely, that the only argument for the claim that the meaning of an expression determines its reference rests upon it. 42 What obscures this point, I think, is Dummetts acceptance of a broadly Davidsonian, interpretive philosophy of mind. If one puts the two views together, as it were, use will be described in terms that do not make use of notions like belief, but instead incorporate the interpretive story about what belief is. Nonetheless, even if one does accept such a view, it is worth separating it out from the view about how use should be characterized: The two are quite independent.

21

Richard G. Heck, Jr. Use and Meaning 4 Closing: The Epistemology of Understanding and the Priority of Thought

spects exactly like ours,43 not just actually but counterfactually, too: Whatever semantic differences there might be between the languages then could never be revealed in anything we could glean from the Martians speech. A Gricean could reply that the differences would yet consist in differences between the contents of the Martians thoughts and ours, and that it might still be possible to give an account of that difference.44 But, by hypothesis, the differences between the contents of their thoughts and ours would be epiphenomenal, as far as their speech is concerned, and Dummett regards that situation as intolerable. Ordinarily, we take ourselves to know, on the basis of how someone uses language, what she means by her words. And yet, in this case, no amount of ordinary probing could reveal our mistaken assumption that the Martians mean what we do. Are we not forced to conclude that the grounds on which we ordinarily judge what someone means are insufcient? or, alternatively, that faith is required if we are to believe that we communicate with one another (Dummett, 1993a, p. 177)? Better, says Dummett, to accept that meaning is determined by use, in his sense. McDowells view is that such arguments are what underlie Dummetts rejection of meaning-laden notions of use. But to say so is to misidentify their target. | The argument we just considered is directed not against meaning-laden notions 550 of use but against content-laden notions of use, in particular, against the claim that the right way to specify the Use-Meaning Thesis is in terms of a Gricean notion of use:45 The target, that is to say, is the thesis that thought is prior to language. It is in response to this argumentperhaps surprisingly, an argument for a claim with which he actually agreesthat McDowells epistemological observations are properly deployed. The case of the Martians is indeed analogous, in relevant reIt wont really be exactly like ours, of course: there may be vocabulary differences, and externalist constraints might imply that their word water means twater. But these kinds of differences are presumably ones that could become apparent. 44 Another reply would be that, even on a Gricean view, the case is impossiblei.e., that, though meaning does not consist in use, described in Dummettian terms, differences in meaning nevertheless will always lead to potential differences in use. Note that this reply in effect concedes that meaning is, in some sense, determined by Dummettian use, a claim that may still be strong enough for the anti-realist challenge to get off the ground. In any event, it is this weaker claim for which I would suggest the name The Manifestation Constraint, familiar from discussions of Dummetts writings. A speakers linguistic capacities manifest what she means in the sense that her having those capacities demands explanation in terms of her having particular semantic knowledge so that, even if her meaning what she does by her words is to be explained in terms of a prior capacity for thought, what she means will still be xed by the use she makes of her language. 45 The quotation at the end of the last paragraph, for example, occurs in a discussion of the view that understanding consists in tacit knowledge of a theory of truth (Dummett, 1993a, pp. 17481). More specically, the target is Chomskys view that, as Dummett puts it, language is primarily a vehicle of thought, and only secondarily an instrument of communication (Dummett, 1993a, p. 176).
43

22

Richard G. Heck, Jr.

Use and Meaning References

spects, to the argument from illusion in the philosophy of perception. And we may follow now familiar responses to the argument from illusion and deny that the mere possibility that there could be such a language, one we could not distinguish from English on the basis of the evidence upon which we ordinarily base our beliefs, shows that our ordinary beliefs about what utterances mean do not constitute knowledge (McDowell, 1998b, p. 244). I think that response exactly right. In any event, it is Dummetts rejection of content-laden notions of use that is epistemologically driven, not his rejection of meaning-laden ones.46

References
Chomsky, N. (1980). Rules and Representations. New York, Columbia University Press. (1988). Language and Problems of Knowledge. Cambridge MA, MIT Press. Davidson, D. (1984a). Inquiries Into Truth and Interpretation. Oxford, Clarendon Press. (1984b). Radical interpretation, in Davidson 1984a, 125139. (1984c). Truth and meaning, in Davidson 1984a, 1736. Davies, M. (1987). Tacit knowledge and semantic theory: Can a ve per cent difference matter?, Mind 44162. Dummett, M. (1987). Reply to John McDowell, in B. Taylor (ed.), Michael Dummettt: Contributions to Philosophy. Dordrecht, Springer, 25368.
Thanks to Alex George, Steve Gross, Chris Peacocke, Ian Proops, Jim Pryor, Alison Simmons, Sanford Shieh, and Crispin Wright for helpful comments on earlier drafts of this paper. Thanks to Jim Higginbotham, Sally Sedgwick, Jason Stanley, and Jamie Tappenden for discussions that contributed to its development. I beneted from the opportunity to present this material in a graduate seminar on The Logical Basis of Metaphysics, in the Autumn of 1994 (and in various undergraduate courses on the philosophy of language taught since then). Thanks to all the participants for their comments, but especially to Michael Glanzberg, Ian Proops (again), and Jamie Tappenden. The paper was also read to a colloquium at MIT, in January 1997. Thanks to all who were present for helpful questions, but especially to Alex Byrne, Vann McGee, and Bob Stalnaker. When I rst wrote this paper, I found Dummetts position, as I interpret it, more attractive than I do now. This change of view left this paper unpublished for a long time. But various people suggested to me, over the years, that its exposition of Dummett was helpful. Others have even found the position it offers Dummett attractive. Thanks to all of them, especially Ian Proops (yet again) and Michael Rescorla, for their encouragement. Without it, this paper never would have been published. Finally, thanks to Michael Dummett, for his teaching, when I was studying for my B.Phil. in Oxford, and for his writings, which have been an inspiration.
46

23

Richard G. Heck, Jr.

Use and Meaning References

(1991). The Logical Basis of Metaphysics. Cambridge MA, Harvard University Press. (1993a). Language and communication, in Dummett 1993b, 166187. (1993b). The Seas of Language. Oxford, Clarendon Press. (1993c). Truth and meaning, in Dummett 1993b, 147165. (1993d). What do I know when I know a language?, in Dummett 1993b, 94105. (1993e). What is a theory of meaning? (I), in Dummett 1993b, 133. Etchemendy, J. (1988). Tarski on truth and logical consequence, Journal of Symbolic Logic 53: 5179. Evans, G. (1985). Semantic theory and tacit knowledge, in Collected Papers. Oxford, Clarendon Press, 322342. Heck, R. G. (1995). The sense of communication, Mind 104: 79106. (1997). Tarski, truth, and semantics, Philosophical Review 106: 53354. (2000). Non-conceptual content and the space of reasons, Philosophical Review 109: 483523. (2004). Truth and disquotation, Synthese 142: 31752. (2005). Reason and language, in C. MacDonald and G. MacDonald (eds.), McDowell and His Critics. Oxford, Blackwells, 2245. Higginbotham, J. (1988). Is semantics necessary?, Proceedings of the Aristotelian Society 87: 219241. (1992). Truth and understanding, Philosophical Studies 65: 316. (1994). Priorities in the philosophy of thought, Proceedings of the Aristotelian Society sup vol 64: 85106. (1995). Sense and Syntax. Oxford, Oxford University Press. Larson, R. and Segal, G. (1995). Knowledge of Meaning. Cambridge MA, MIT Press. McDowell, J. (1996). Mind and World. Cambridge MA, Harvard University Press. 24

Richard G. Heck, Jr.

Use and Meaning References

(1998a). Another plea for modesty, in McDowell 1998d, 10831. (1998b). Anti-realism and the epistemology of understanding, in McDowell 1998d, 314343. (1998c). In defense of modesty, in McDowell 1998d, 87107. (1998d). Meaning, Knowledge, and Reality. Cambridge MA, Harvard University Press. (1998e). Truth-conditions, bivalence, and vericationism, in McDowell 1998d, 328. Peacocke, C. (1992). A Study of Concepts. Cambridge MA, MIT Press. (1997). Concepts without words, in R. Heck (ed.), Language, Thought, and Logic: Essays in Honour of Michael Dummett. Oxford, Clarendon Press, 133. Rumtt, I. (1995). Truth-conditions and communication, Mind 104: 827862. Segal, G. (1994). Priorities in the philosophy of thought, Proceedings of the Aristotelian Society sup vol 68: 10730. Shieh, S. (1998). On the conceptual foundations of anti-realism, Synthese 115: 3370. Soames, S. (1984). What is a theory of truth?, Journal of Philosophy 81: 41129. Wright, C. (1993). Theories of meaning and speakers knowledge, in Realism, Meaning, and Truth. Cambridge, Blackwell, 204238. 2d ed.

25